Chicago E-scooter Pilot Program

The city of Chicago ran a pilot program in 2020 (from August to December) where 10,000 ride share e-scooters were permitted to operate within allocated community areas. 3 vendors participated in the trail each supplying 1/3 of the e-scooters.

The below map displays the area of Chicago involed within the pilot program. The areas in grey are the community areas participating in the program. The areas in red are communities and high traffic regions which e-scooter use is prohibited. The area within the blue line outlines the equity priority area. The equity priority area "covers neighborhoods where residents face systemic disadvantages following generations of underinvestment and inequitable access to resources". As part of the escooter program at least 50% of the vendors escooters must be deployed within this area. More information regarding the escooter program can be found here.

1. E-scooter Data Exploration

1.1 Importing Data

The E-ScooterTrips-_2020 data set contains trip data collected during the trial. The data was comprised of the following attributes:

The trip mean distance is 2.91 kms with a minimum trip of 1 m and a maximum tip of 50 km.

The trip distance data has a coefficient of variation of 1.28. This value is large and may be effected by outliers.

The trip mean duration is 16 mins and 36 secs, with a minimum trip of 0 seconds and a maximum trip of 56 hours 42 mins 36 secs.

The trip duration data has a coefficient of variation of 1.36. This value is large and may be effected by outliers.

1.2 Identifying and Addressing Outliers

The descriptive statistics indicate that a number of outliers may be present within the data which are influencing the results. A cut off range was established as 3 standard deviations from the mean and these values were dropped from the data.

Also, trips that were less then 5 seconds in duration were dropped.

Upon the removal of outliers the trip mean distance is 2.5km with a min of 1 m and max of 14.06km.

The coefficient of variation was reduced to 0.99.

The trip mean duration is 15mins 2 secs with a minimum of 6 seconds and maximum of 2 hrs 48 mins and 33 seconds.

The coefficient of variantion was reduced to 1.16.

1.3 Investigating Null Values

A number of null values exist within the data set and these null values were investigated.

A number of start community area name and end community area name data fields are null.

The community areas with the most start and end community name null values border the exlusion zones for the escooter pilot program. It can be infered that these values are null as users are starting or finishing their journies outside of the permitted pilot program zone. As such the GPS data for these values have been exluded from the data.

The largest number of nulls for both start and end community name feild was in the Austin. Further investigation is recomended into why more more trips start or end outside of the allocated escooter area within the Austin community area.

The Lime vendor make up the majority of trips with a start or end null value for the community name field. This may be due higher distribution of this vendors escooters within areas such as Austin on the pilot program border or an issue with how this vendor is communicating the location of exlusion zones to the user. Investigation should be undertaken to understand why users are taking Lime scooters outside of the permitted usage zones.

1.4 Distribution of Trip Volume by Starting Location Between Community Areas

The data was investigated to understand the distribution of trip starting locations across the city.

The mean number of trips starting within a community area is 8,004.64. The standard deviation is large suggesting a significant range of values between the number of starting trips within community areas.

The histogram shows the a significant right skew with only 3 community areas having total trips starting within that community numbering above 50,000. The majority of community areas was a starting location for less then 20,000 trips.

The most popular starting point locations for escooter trips was the communities of Lake View, Lincoln Park and West Town. These community areas are located to the north of the city and are the location of a number of tourist attractions such as North Avenue Beach, Lincoln park and zoo, Diversey Habour and a number of large parks.

The least popular starting locations were Ohare, Edison Park, Burnside and the Loop. Edison park and Burnside are geographically small communities which may explain the lack of trips starting from these communities. Ohare and the Loop are zones of exlusion for the program as outlined by the E-Scooter Pilot Map.

The data was investigated to understand the distribution trip end locations.

1.5 Distribution of Trip Volume by End Location Between Community Areas

The descriptive statistics and the distribution of trip end volume by community area is similar to that of the trip start data. This would suggest that many of the trips undertaken by e-scooter are localised to within the community area. This can be further explored by comparing the difference between trip start and trip end volume within each community area.

1.6 Migration of E-scooter Trips between Community Areas

The data was explored to identify the difference between the number of trips started within each community area to the number of trips which ended within that community area. This will give an indication if e-scooter trips are migrating towards or away from specific community areas.

A negative difference between end and start trip volume indicates users are starting a trip in that community area and ending their trip in another community area. The communities with the largest negative numbers include Lincoln Park, Lake View, Hyde Park and Uptown. The mapa above shows that these communiy areas are surrounded by community areas with posistive e-scooter migration. This would indicate that users are moving from these listed communities to the surrounding communities.

It should be noted that the difference between start and end trip numbers are very small relative to the total number of trips taken within that community area. This would indicate that the majority of e-scooter trips start and finish within the same community area.

1.7 E-scooter Usage During The Program by Number of Trips

The data was investigated to understand the pattern of e-scooter usage over time. This involved exploring the number of trips during the time period, the total distances traveled and the total time duration.

The e-scooter program lasted for 123 days total with an average of 5,018.14 trips taken per day.

The number of trips per day can vary significantly from day to day. There was a initial peak in number of trips within the first couple of weeks of the pilot program. These numbers seemed reasonably consistent for the first month. From October 2020 the number of trips began to decline and continued to do so until the end of the pilot program.

The most e-scooter trips occurred on a Saturday followed by Friday and Sunday. The higher usage of E-scooters on weekends would suggest that this mode of transport is being utilised for recreational purposes rather than the daily commute to and from work.

E-scooter usage builds from early morning (5am) to a peak of 65,959 trips at 5pm, then drops off. From the hours of 11pm to 4am e-scooters are not used. Relatively low usage during the morning peak hour times suggests that the e-scooters are not being used for travel to work. Taking into consideration that usage is significantly higher on the weekends and usage is higher during the afternoon/evening it would suggest that the scooters were used primarily for recreational transport.

1.8 E-scooter Usage During the Program by Total Distance Travelled and Total Duration of Trips

The average total distance traveled per day is 12,765.58 km with a min of 1,220.07 km and a max of 34,686.46 km.

The average total duartion of trips per day is 1257 hrs 40 min 41 sec with a min of 112 hrs 44 min 33 sec and a max of 3859 hrs 22 min 27 sec.

The total distance travelled per day and total trip duration per day follows a similar pattern to the number of trips taken per day. The variability between days is large, the highest values are observed during the begining of the pilot period and remain relatively level during September after which they gradually decline for the remaining period of the pilot program.

1.9 Vendor Investigation

The data was investigated to understand the pattern of usage between vendors.

Lime was the most popular vendor with 278,899 trips. The vendors Spin and Bird had a similar number of trips.

The map shows that the Lime vendor is the most popular vendor for most communities across the city. The spin vendor is popular for a collection of communities to the north east of the city and includes the community with the most trips, Lake View. The bird vendor is popular for a number of communities which are sporadically spread across the city.

During the very early stages of the pilot program, bird was the most popular vendor however lime quickly took over and stayed the most popular throughout most of the pilot program. The bird and spin vendors remained similar in popularity throughout the program. Trip numbers fell for all vendors at similar rates from October to the end of the program.

1.10 Summary

Upon the removal of outliers from the e-scooter trip data it was found a total of 617, 231 trips were completed over the 123 days of the pilot program. The average trip distance was 2.5km and the average trip duration lasted for 15 minutes and 2 seconds.

Several null values were noticed within the start and end community area name attributes. Upon investigation it was concluded that the most likely reason for the null values was that the scooter had started or completed a trip outside the community areas participating within the program. The largest number of null values were recorded on the border communities. The community of Austin had the largest number of null values. The lime vendor had a disproportionate number of null values in comparison to the spin and bird vendors. It is suggested that the way in which the lime vendor communicates where a user can and cannot use the scooter is evaluated.

It was found that the starting location of trips were not evenly distributed throughout the communities. The communities of Lake View, Lincoln Park and West Town had a larger number of trips compared to the other communities. These communities are the location of several popular recreational hotspots within Chicago including a number of large parks, beaches and the zoo. The communities where the least number of trips started was Ohare and Edison Park. This can be explained as a large portion of these communities make up the e-scooter exclusion zone.

The number of trips ending within a community followed a near identical distribution to the number of trips starting in a community. This would suggest that most trips are localised to within a community area. This is further supported by the migration pattern data which showed only a small number of trips started in one community and ended in another. The community with the largest negative trip migration Lincoln Park. As surrounding communities had positive migration users are starting trips within Lincoln Park and travelling to surrounding suburbs. Over the 123 day period of the program it was found that the number of trips taken peaked early and remained relatively stable throughout September. However from October onwards the number of trips began to decline at a steady rate until the program concluded. The average number of trips taken per day was 5,018.14.

A difference was observed within the number of trips by day of the week. Saturday had the largest number of trips taken followed by Friday and Sunday. The least number of trips taken were taken on Mondays and Tuesdays. This suggests that the scooters are being used for recreational travel rather than everyday commuting. This is supported by the most popular hour of the day that trips are taken. The most popular time for a trip to be taken was in the afternoon/evening between 5pm and 6pm. Of note was the small number of trips taken during the typical morning peak hour times.

It was found that Lime was the most popular vendor completing the largest number of trips. The vendors spin and bird completed a similar number of trips.